BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260626T134757Z
LOCATION:Bldg. 8 - B 102
DTSTART;TZID=Europe/Stockholm:20260701T103000
DTEND;TZID=Europe/Stockholm:20260701T110000
UID:submissions.pasc-conference.org_PASC26_sess157_msa181@linklings.com
SUMMARY:pyiron – A Workflow Framework for Trustworthy Agentic Workflows in
  Materials Science
DESCRIPTION:Jan Janssen (Max-Planck-Insitute for Sustainable Materials)\n\
 nThe hierarchical nature of materials requires simulation approaches that 
 couple methods across disciplines and length scales. Integrating heterogen
 eous simulation codes poses significant interoperability challenges due to
  incompatible units, file formats, and data structures. The pyiron workflo
 w framework addresses these challenges by providing Python interfaces that
  represent simulation codes as standardized objects, enabling reproducible
 , data-driven materials science workflows while abstracting technical comp
 lexity.\n\nBuilding on this foundation, we integrate agentic AI capabiliti
 es through a team of large language model (LLM) agents that autonomously c
 onstruct and execute simulation workflows. The agents plan simulation camp
 aigns, perform convergence tests, and analyze results to generate new insi
 ghts beyond existing high-throughput databases. Structured information sha
 ring between agents through a shared workflow “canvas” significantly impro
 ves coordination and performance. Benchmarks show that the agent system ca
 n reach simulation accuracy comparable to domain experts in density functi
 onal theory calculations by systematically studying basis-set convergence.
 \n\nWe demonstrate how these agentic workflows scale to high-throughput sc
 reening across the periodic table and integrate with materials acceleratio
 n platforms that couple simulations with experiments. This work highlights
  how combining agent-based reasoning with HPC workflow systems enables tru
 stworthy, reproducible, and scalable discovery pipelines for computational
  materials science.\n\nDomain: Chemistry and Materials, Climate, Weather, 
 and Earth Sciences, Engineering, Physics\n\nSession Chair: Jan Janssen (Ma
 x Planck Institute for Sustainable Materials)\n\n
END:VEVENT
END:VCALENDAR